NATURAL LANGUAGE PROCESSING - Mandatory IDMP Compliance is Almost Here - How NLP Can Help

INTRODUCTION

The Identification of Medicinal Products (IDMP) is a set of five standards developed by the International Organization for Standardization (ISO) to help streamline and improve the safety of pharmaceutical operations across the entire drug development cycle. The overarching goal is to enhance patient safety through an improvement in the consistency, accuracy, and speed of shar­ing adverse events (AE) and safety signal reporting. These stan­dards were originally developed in 2012 and have been periodically updated since. Now, 11 years following develop­ment, the IDMP standards are set to become mandatory this year.

THE IDMP FRAMEWORK

The five standards that define the IDMP framework are de­scribed in Table 1. IDMP specifies the use of standardized defini­tions across global regulatory and health authorities for the descriptions and identification of medicinal products, with the purpose of facilitating reliable and consistent exchange of me­dicinal product information. IDMP impacts many aspects of the drug lifecycle, including marketing, data management, regulatory operations and affairs, and pharmacovigilance. The five stan­dards that make up IDMP look to homogenize the descriptions of marketed medicinal products around the world, with the main end goal being improved patient safety. These standards have been found to be impactful in practice; the EU commission esti­mates that improved pharmacovigilance regulations will save up to 5,910 lives per year.

TIMELINE

The timeline for implementation differs by health authority. The European Medicines Agency (EMA) is set to be the first health authority to mandate IDMP compliance in 2023 and recently re­leased its guidelines for implementation. Other global health au­thorities across the world will soon implement IDMP, and the FDA is one of the next in line. While these standards will have a large, positive impact in the space, pharma organizations need to pre­pare for the challenges that come with the transition.

DATA CONSIDERATIONS FOR IDMP

Medicine safety is a global issue; there is an urgent need for accurate medicinal data, but this key data does not always trans­fer well. With IDMP, consistent data for drug processes will strengthen drug safety and regulations across borders. It will also be extremely beneficial in times of medicine shortages – the har­monization of drug descriptions can assist in finding a different, suitable drug option quickly.

IDMP compliance will take resources, time, and investment. IDMP-related data is captured in unstructured data containers, and these documents may have different formats, different lan­guages, or different phrasing used for the same terminology. Dif­ferent authors tend to have individualized styles, and these factors create difficulties in manual search when capturing data entries from both internal and external document sources for IDMP. Key­word searches and manual curation are not only slow, but they are tedious, limited, and prone to errors. The capture of 300-2000 data entities per product demands major investment.

AI & NLP FOR IDMP COMPLIANCE

This is where Natural language pro­cessing (NLP) emerges as an asset. NLP is a form of artificial intelligence (AI) that has the ability to understand human language, both written and spoken. NLP can scan through documents or data to process and organize the content. NLP reduces time, effort, and the overuse of resources that emerge with extracting, structuring, and standardizing required data elements from unstructured IDMP text documents.

Around 80% of biomedical data is locked in unstructured text – without access to all text, information and insights are missed and lost. NLP transforms text into structured information and metadata – not only replacing the need for manual key­word search and cutting down on the time required, but divulging more information as well. In relation to IDMP, the documents and rich text data would be in the form of eCTD documents, regulatory models, field notes, or even social media posts. NLP can sift through documents and find different words, expressions, or grammar with the same meaning, or the same word with a different context. The information of the specific query posed is then highlighted and surfaced in a structured format. This reduces the efforts needed to abide by the IDMP standards and assists in a smooth transition to the new requirements.

IDMP PREPARATION TAKEAWAYS

The best way to prepare for IDMP im­plementation is by knowing exactly what is required to achieve such implementation and how to get there. With the US sched­uled to implement IDMP this year and the EU in the process of implementation, IDMP is no longer just a distant requirement. It is happening now, and all organizations should be prepared for mandatory use this year.

Simon Johns has more than 25 years of experience supporting customer projects across all stages of drug development and the full product lifecycle. As Director of Medical Information (MI) and Marketed Product Safety at IQVIA, he has been managing global MI projects focused on process optimization and technology enablement that drive enhanced efficiency and customer engagement. He is a member of the European DIA Medical Information and Communications Training Team, advising pharmaceutical companies on best industry practices, innovation, and automation. He speaks regularly on topics ranging from implementing suitable technologies and innovations to optimize MI to the benefits of integrating MI and pharmacovigilance to increase compliance and product value, leveraging IQVIA’s Local Affiliate Product Services (LAPS), which provide full support for MI and local country pharmacovigilance requirements.